What accuracy can be expected?

Using a 24" DRO I find that my machine varies by as much as 0.009" when I traverse in one inch increments (X or Y). It repeats well (the error is the same on multiple tests), but doesn’t achieve the accuracy I would like at all distances from home. The machine has two years of use and I just changed the belts in hopes of improving the accuracy, but that did not make any difference. If a servo is bad, you would think it would not repeat. I am stumped. Many test distances are within 0.003, which is fine. The error is random and does not build, so calibration will not solve the problem. Any thoughts or suggestions would be appreciated.

Account for backlash?

A properly tuned Xcarve is expected to provide an accuracy = 0.1-0.15mm (IIRC)

If you want that level of accuracy, may I ask what you are trying to machine? Metal?

I am a luthier and need to make parts with cylindrical projections that fit into a hole pattern. At the moment, my cylinders are out of round, as are my holes. It seems that the machine was more accurate for the first year and a half. To get things to fit, I am needing to make the cylinders 0.010 undersize, which is more that desired. I used to be able to go almost size on size and parts came together with a tight slide fit.

Yes and mine was well within this range when new. Being 0.009" out is problematic.

Did you re-tighten your belts recently? This could have an impact on precision. Belts have to be just tight enough to play a chord.

Yes, the belts are tight. I build instruments, so listening for the right pitch is natural. What is puzzling, is the fact that it can repeat, so that for instance, if the test location at 7.000" is off by +0.009 on one trial, it will be the same +0.009 off on repeated trials. The test location might be exactly at 12.000 on each test. The error is random, but repeats exactly for subsequent tests. That tells me it isn’t lost motion. Plus I am doing all tests in one direction, so lost motion should’t be a factor. I am tempted to buy a new drive and see if that improves things.

1 Like

Did you compensate for belt stretch?

Yes, the overall error at 22" has been compensated for. It’s the error at points between zero and 22" that are problematic.

1 Like

I dont think its due to lost motion either, its 99% likely to be something mechanical.
V-Wheels wear over time, your information indicate that the discrepancy have increased over time?

When cylindrical carves are out of round the main reason for this is difference in rigidity between X/Y axis.
The sideway cutting force allow for more movement in one direction vs the other.
There have been cases where set screws are almost tight enough, causing the pulley to slide ever so slightly on the motor shaft, this can also cause similar symptoms. But that would be revealed if tested for backlash specifically.

Thats a good thought. I will set up a test with a known force in the X and then the Y directions and see if there is any difference in the deflections.

I should have given you my website so you can see what I typically make: stephenslutherie.com

very cool stuff!
Unfortunately I can’t add much to the discussion that hasn’t been said already. Good luck anyways :wink:

I added a lead screw kit to two of my Xcarves, Similiar to you, I had a bit of constant error on my travels upon setup and I just tweaked the steps/mm in the GRBL settings to make the machine precise. I dumped the belts because I am engraving and de-burring precision parts at my machine shop with the Xcarves and all I got was headaches. Look at your steps per millimeter (settings $100, $101,$102) and using a calculator multiply it by (1/1.009) or 0.9910802775024777 and write the new value to GRBL for each axis with the error.

The other item I discovered was the Y motors wouldn’t always be timed right after some use. Maybe that was due to running over debris on the V wheels, etc etc… What I got was a twisted gantry. Instead of doing stuff like clamping the belts while the motors were all holding position I went to lead screw. Occasionally an error pops up, but I have measurements on each side of the gantry to reset it manually.

I’ll blame the belts. !

The tests did show that the machine is more rigid along the Y axis than the X by about 20%. The cylinders I make are larger (.004" on a .625 diameter) along the X axis, which corresponds to the lack of stiffness on that axis.

I may be expecting too much from the belts. I love the simplicity and low cost, but it is probably not reasonable to expect that a belt will not have error greater than 0.003" along it’s length. It is calibrated so that it reads exactly right at 20.000 from home. It’s just not within 0.003 at EVERY point between home and 20.000. I don’t know how accurate a cheap lead screw is, but I would think it is within my 0.003 criteria. I will have to consider the lead screws.

I use gage blocks on both Y axis when I power up and that seems to be good enough for repeatability. I can usually drop right in to a previously carved spot with no discernible positioning error.

I am also considering redoing my fixtures so that when I am making two parts that need to fit together precisely, I am carving at the exact same point on the spoil board. That way machine error of the type I am experiencing will not matter.

It’s a challenge to get a $1100 cnc router to do very precise work, but I have been impressed with the X-Carve’s ability to make functional components if you work with it’s limitations.

Your idea to place the sister parts in one location is perfect. THAT is the way to use a cheap machine! Bonus point for your skilled thought.

The tests did show that the machine is more rigid along the Y axis than the X by about 20%. The cylinders I make are larger (.004" on a .625 diameter) along the X axis, which corresponds to the lack of stiffness on that axis.

Hi Bob, Tell me about your spindle and Z assembly, as I noticed more play in my dewalt over time, (side to side play). I cannot speak to your specific problem but I can tell you that I changed all my v-wheels (12) in the entire carriage at the same time I replaced my spindle. Basically I took the entire carriage apart and then put it back together. I found the v-wheels to be the biggest issue in my case. I have good dust collection, but the left side of the carriage had twice the build up of gunk, deep into the v-wheel cavity. Even after cleaning I still had play, BUT after replacing all at the same time (because I surmised the tolerance between new and old v-wheels or bearings might differ) My issue was resolved, which was also getting a more accurate circle. I did not measure these tolerances to the accuracy you are, I just measured completed carves and used the inlay feature in easel, rotating the inner circle piece and looking for any gaps. Best of luck with this, and like the work on your site too!

After a lot of testing and measuring, I have concluded that the major error I am experiencing is in the belt itself. As a result, it repeats every time I take measurements. Since there is nothing I can do about belt error, I am resigned to live with it. The belt is actually more accurate than the rolled ball screws, so unless you go all the way to ground screws, you might as well work around the belt error as best you can. I suspect most machines are experiencing the same issue- the owners just don’t have the equipment to measure the error. The time the error comes into play is when one part has to fit into another in a sort of inlay fashion. If I locate the two parts in the same spot on the table when I make them, I can achieve accurate enough results. In the end, I guess that’s all that matters.

Thanks to all who weighed in on this thread.